wyn314/hunspeech

emSpeech

emSpeech is the speech module of the e-magyar language processing system.

Speech activity detection (emSad)

The speech activity detection module wraps the SHOUT.

Speaker diarization (emDia)

The speaker diarization module wraps the 'shout_segment' and 'shout_cluster' programs of SHOUT, with added audio conversion using the SoX (Sound eXchange) utility. For details on the usage run

python speaker_diarization/em-dia.py --help

Installation

1. Install SHOUT. Instructions can be found here.

2. Install sox. sox should be available on most Linux distribution (e.g. apt-get install sox).

3. Check out this repository.

4. Set the SHOUT_DIR environmental variable to the directory where you isntalled SHOUT:

   export SHOUT_DIR=/path/to/shout

This will be the default place to look for the model files SHOUT needs.

Citation

Please cite Marijn Huijbregts's dissertation on SHOUT.